The problem of robot path planning is relatively complex under dynamic environment. During the process of real-time local path planning in local environment within the global path, sudden appearance of obstacles may cause oscillations for mobile robots path planning. If seriously enough, it will result in failures of path planning, and thus the mobile robots are incapable of finishing autonomous navigation. In order to solve this problem, we proposed an algorithm for local path planning and obstacle avoiding based on improved Morphin search tree. Through a multilayer local replanning and by using the corresponding function as a basis for the property evaluation of search tree, a smooth and trackable trajectory with nonholonomic constraints of mobile robots is obtained. This algorithm overcomes the shortcomings of simplex and inflexibility of Morphin algorithm in trajectory searching. The test on a real robot Pioneer3 verifies the correctness and effectiveness of the algorithm.